WIP: fix(vi): pace the 30 FPS cap in the host

TheDevil commented

2026-11-14 23:45:26 +00:00

First-time contributor

Type: fix · Risk: medium · Tested on: Windows, llvm-mingw game-free fixtures · Needs: in-game Windows and Linux pacing validation

Why

Players cannot select a host-paced 30 FPS limit: HOS_FPS_MAX=30 is raised to 60 on main. This extracts and corrects the frame-cap portion of #173; it does not include input, wheels, NR or the patched-runtime bridge. See #306 for the related request for lower frame limits; this change limits game frames, not additional generated frames.

How to reproduce

Seen on: production host code exercised by Windows llvm-mingw fixtures; an in-game reproduction for this revision is still pending.
Settings and switches: HOS_FPS_MAX=30; alternatively the new menu choice, with Game speed set to Per frame.
Steps: run tests/test_frame_rate_cap_30.py --cc clang and tests/test_frame_rate_cap_step.py --cc clang.
Expected: the host, game limiter and vsync period target 30; elapsed game time follows wall time; only one pacing mechanism is used.
Actual: main clamps the environment override to 60 and does not recognize the menu's new -1 choice. The combined PR's separate renderer sleep masked this rather than fixing the host.

What changed

Thirty FPS is now a choice of the existing host pacer. The engine limiter, display period and time-step hooks agree on that rate; the renderer needs no second limiter.

host/nn_vi.c: accept 30 from the menu and environment, keep initialization separate from the valid -1 menu value, keep interval 1 for the 30 Hz period, override Per frame at 30, and restore the normal limiter/step when returning to 60.
gpu/vk/fps_stats.h and the menu: matching 30 FPS graph target, existing 0–6 saved cap values unchanged, explicit help text for the real-time override, translated notices.
Regression fixtures cover 30/60/high-cap transitions, both override routes, single reads of the environment/ini, Per frame, and time-step integration. No per-present getenv or sleep_until is added to the renderer.

Measurements

Verdict: game-free regression checks pass; current-revision in-game A/B measurements are not yet available. The previous combined PR's renderer-sleep measurements are not reused as evidence for this host fix.

Before: main @ 2b1d38c682
After: this PR @ 74f1da1924
Hardware: Windows x64, i7-14700F, 16 GB RAM; GPU/driver not exercised by the game-free fixtures.
Scene: authored option/clock/input fixtures; no gameplay rendered for this revision.
How to reproduce: run python -I -B tests/run_all.py with the official llvm-mingw compiler on PATH; focused commands are listed under Testing.
Method: extracted production functions compiled with llvm-mingw, fake options/clock/guest state, no game or driver calls.

Check	Before: main @ `2b1d38c6`	After: this PR @ `74f1da19`
`HOS_FPS_MAX=30` host rate	60 (clamped by `fps_of`)	30
Menu cap -1 host rate	60 (unknown index)	30
30 FPS native pacing period	no 30 Hz choice	33,333,333 ns, interval 1
Full game-free suite	530 pass, 14 fail, 34 skip	531 pass, same 14 fail, 34 skip

Flags and settings

The menu adds 30 without changing its default of 60. Existing options-file cap indices remain compatible.

Flag	What it is	Before this PR	After this PR	Effect	Helps / doesn't help / untested	Recommendation
frame_rate_cap	Existing menu/ini key	0–6; default 0 (60)	-1 adds 30; default still 0	single host cap; 30 forces real time	useful for a lower render target; in-game timing still needs validation	keep 60 default
HOS_FPS_MAX	Existing run override	30 clamped to 60	30 accepted	overrides the menu; cached once	no promise of minimum FPS	unset by default

Testing

Tested: Windows game-free full suite on both named revisions; focused cap, time-step, graph, environment and translation checks pass. git diff --check passes. Compiler: the installed official nightly llvm-mingw toolchain.

The identical baseline failures are: tests/test_appimage_player_build.py, tests/test_cpm_discovery.py, tests/test_crash_dialog.py, tests/test_env_on.py, tests/test_finish_output.py, tests/test_lift_cleanup_error.py, tests/test_linux_preflight.py, tests/test_make_game_cache.py, tests/test_oot_state_tap.py, tests/test_readback_late.py, tests/test_tested_configurations.py, tests/test_vk_pgo.py, tests/test_vram_budget_row.py, tests/test_walk_key.py. The before/after logs are retained locally; no game files or local paths are uploaded.

Not tested: this revision in a running game, Linux build/runtime, physical displays at different refresh rates, cutscene/audio timing and generated-frame output. WIP intentionally records these gaps rather than reusing older combined-feature results.

Must be tested:

Windows, supported GPU, by hand: stationary gameplay and a cutscene at menu 30 and HOS_FPS_MAX=30; alternate 30/60 twice, verify roughly 33.3/16.7 ms presents and normal game/audio time, including saved Per frame. Record two interleaved runs each at these commits, expecting correct timing/input behavior without regressions (by hand).
Linux, supported Vulkan GPU, by hand: build the host and repeat the same transitions, checking the native interval and absence of doubled pacing, expecting a successful native build and the same behavior (by hand).

Risks

What could break: switching the game's limiter and nominal/timed step at 30 could expose game-specific frame-count assumptions, display pacing or cutscene timing problems. HOS_GAME_INTERVAL, no-pacing and frame-generation combinations still need live runs.

Turning it off: choose an existing cap (60/default) and unset HOS_FPS_MAX; reverting this commit restores the prior selection range.

Compatibility: no save changes, no shader/cache changes, no vk ABI change. Default and saved cap indices 0–6 retain their meanings. New -1 selects 30; older builds fall back to 60 for it. The saved Game speed preference is retained and used again at 60.

Checklist

Everyone

One topic.
The title is a Conventional Commit (type(scope): description) of the status line's type.
Why gives the reason plainly and links the issues that fit (the tracker searched).
Measurements has a verdict, or one line saying why there is nothing to measure.
Testing names what was tested, what wasn't, and what must still be tested, where and how.
Rebased on current main.
python -I -B tests/run_all.py passes (any failure that also fails on main is named).
No game files, dumps, build output or generated files.
third_party/ untouched (a new folder with a PROVENANCE.md instead).
The license is OK: GPL-3.0-or-later, nothing incompatible and nothing decompiled.

When it applies

AI assistance recorded with Assisted-by: trailers.
Numbers from two interleaved in-game runs and their medians: pending the runs above; WIP.
Switches documented and default preserved.
Current-revision screenshots and measured-run docs/tested/ fragment: pending live validation; WIP.
Game-free regression added; changed source-reading checks updated without weakening their assertions.
Host build/runtime on both Windows and Linux: game-free Windows fixtures passed; Linux pending.
Docs updated.

Upstream refresh

Current head: a9c0551156ae2043f441b4cfee05f1784fd58ed5. Main snapshot 6c9edffe is incorporated by merge, preserving published history. 5 focused checks passed after the merge; zero failures/skips. Earlier full-suite measurements remain tied to their named feature commits above. The literal rebase checklist remains unticked because history was merged, not rewritten. Fresh game/Linux checks remain WIP.

**Type:** fix · **Risk:** medium · **Tested on:** Windows, llvm-mingw game-free fixtures · **Needs:** in-game Windows and Linux pacing validation ## Why Players cannot select a host-paced 30 FPS limit: `HOS_FPS_MAX=30` is raised to 60 on main. This extracts and corrects the frame-cap portion of https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/173; it does not include input, wheels, NR or the patched-runtime bridge. See #306 for the related request for lower frame limits; this change limits game frames, not additional generated frames. ## How to reproduce **Seen on:** production host code exercised by Windows llvm-mingw fixtures; an in-game reproduction for this revision is still pending. **Settings and switches:** `HOS_FPS_MAX=30`; alternatively the new menu choice, with Game speed set to Per frame. **Steps:** run `tests/test_frame_rate_cap_30.py --cc clang` and `tests/test_frame_rate_cap_step.py --cc clang`. **Expected:** the host, game limiter and vsync period target 30; elapsed game time follows wall time; only one pacing mechanism is used. **Actual:** main clamps the environment override to 60 and does not recognize the menu's new -1 choice. The combined PR's separate renderer sleep masked this rather than fixing the host. ## What changed Thirty FPS is now a choice of the existing host pacer. The engine limiter, display period and time-step hooks agree on that rate; the renderer needs no second limiter. - `host/nn_vi.c`: accept 30 from the menu and environment, keep initialization separate from the valid -1 menu value, keep interval 1 for the 30 Hz period, override Per frame at 30, and restore the normal limiter/step when returning to 60. - `gpu/vk/fps_stats.h` and the menu: matching 30 FPS graph target, existing 0–6 saved cap values unchanged, explicit help text for the real-time override, translated notices. - Regression fixtures cover 30/60/high-cap transitions, both override routes, single reads of the environment/ini, Per frame, and time-step integration. No per-present `getenv` or `sleep_until` is added to the renderer. ## Measurements **Verdict:** game-free regression checks pass; current-revision in-game A/B measurements are not yet available. The previous combined PR's renderer-sleep measurements are not reused as evidence for this host fix. Before: main @ 2b1d38c682baaa4d52f2cacc75fffb9505c1f34f After: this PR @ 74f1da1924731d150c59f7de0270d3742a05681a Hardware: Windows x64, i7-14700F, 16 GB RAM; GPU/driver not exercised by the game-free fixtures. Scene: authored option/clock/input fixtures; no gameplay rendered for this revision. How to reproduce: run `python -I -B tests/run_all.py` with the official llvm-mingw compiler on PATH; focused commands are listed under Testing. Method: extracted production functions compiled with llvm-mingw, fake options/clock/guest state, no game or driver calls. | Check | Before: main @ 2b1d38c6 | After: this PR @ 74f1da19 | |---|---|---| | `HOS_FPS_MAX=30` host rate | 60 (clamped by `fps_of`) | 30 | | Menu cap -1 host rate | 60 (unknown index) | 30 | | 30 FPS native pacing period | no 30 Hz choice | 33,333,333 ns, interval 1 | | Full game-free suite | 530 pass, 14 fail, 34 skip | 531 pass, same 14 fail, 34 skip | ### Flags and settings The menu adds 30 without changing its default of 60. Existing options-file cap indices remain compatible. | Flag | What it is | Before this PR | After this PR | Effect | Helps / doesn't help / untested | Recommendation | |---|---|---|---|---|---|---| | frame_rate_cap | Existing menu/ini key | 0–6; default 0 (60) | -1 adds 30; default still 0 | single host cap; 30 forces real time | useful for a lower render target; in-game timing still needs validation | keep 60 default | | HOS_FPS_MAX | Existing run override | 30 clamped to 60 | 30 accepted | overrides the menu; cached once | no promise of minimum FPS | unset by default | ## Testing **Tested:** Windows game-free full suite on both named revisions; focused cap, time-step, graph, environment and translation checks pass. `git diff --check` passes. Compiler: the installed official nightly llvm-mingw toolchain. The identical baseline failures are: `tests/test_appimage_player_build.py`, `tests/test_cpm_discovery.py`, `tests/test_crash_dialog.py`, `tests/test_env_on.py`, `tests/test_finish_output.py`, `tests/test_lift_cleanup_error.py`, `tests/test_linux_preflight.py`, `tests/test_make_game_cache.py`, `tests/test_oot_state_tap.py`, `tests/test_readback_late.py`, `tests/test_tested_configurations.py`, `tests/test_vk_pgo.py`, `tests/test_vram_budget_row.py`, `tests/test_walk_key.py`. The before/after logs are retained locally; no game files or local paths are uploaded. **Not tested:** this revision in a running game, Linux build/runtime, physical displays at different refresh rates, cutscene/audio timing and generated-frame output. WIP intentionally records these gaps rather than reusing older combined-feature results. **Must be tested:** - Windows, supported GPU, by hand: stationary gameplay and a cutscene at menu 30 and `HOS_FPS_MAX=30`; alternate 30/60 twice, verify roughly 33.3/16.7 ms presents and normal game/audio time, including saved Per frame. Record two interleaved runs each at these commits, expecting correct timing/input behavior without regressions (by hand). - Linux, supported Vulkan GPU, by hand: build the host and repeat the same transitions, checking the native interval and absence of doubled pacing, expecting a successful native build and the same behavior (by hand). ## Risks **What could break:** switching the game's limiter and nominal/timed step at 30 could expose game-specific frame-count assumptions, display pacing or cutscene timing problems. `HOS_GAME_INTERVAL`, no-pacing and frame-generation combinations still need live runs. **Turning it off:** choose an existing cap (60/default) and unset `HOS_FPS_MAX`; reverting this commit restores the prior selection range. **Compatibility:** no save changes, no shader/cache changes, no vk ABI change. Default and saved cap indices 0–6 retain their meanings. New -1 selects 30; older builds fall back to 60 for it. The saved Game speed preference is retained and used again at 60. ## Checklist ### Everyone - [x] One topic. - [x] The title is a Conventional Commit (`type(scope): description`) of the status line's type. - [x] Why gives the reason plainly and links the issues that fit (the tracker searched). - [x] Measurements has a verdict, or one line saying why there is nothing to measure. - [x] Testing names what was tested, what wasn't, and what must still be tested, where and how. - [ ] Rebased on current `main`. - [x] `python -I -B tests/run_all.py` passes (any failure that also fails on `main` is named). - [x] No game files, dumps, build output or generated files. - [x] `third_party/` untouched (a new folder with a `PROVENANCE.md` instead). - [x] The license is OK: GPL-3.0-or-later, nothing incompatible and nothing decompiled. ### When it applies - [x] AI assistance recorded with `Assisted-by:` trailers. - [ ] Numbers from two interleaved in-game runs and their medians: pending the runs above; WIP. - [x] Switches documented and default preserved. - [ ] Current-revision screenshots and measured-run `docs/tested/` fragment: pending live validation; WIP. - [x] Game-free regression added; changed source-reading checks updated without weakening their assertions. - [ ] Host build/runtime on both Windows and Linux: game-free Windows fixtures passed; Linux pending. - [x] Docs updated. ## Upstream refresh Current head: `a9c0551156ae2043f441b4cfee05f1784fd58ed5`. Main snapshot `6c9edffe` is incorporated by merge, preserving published history. 5 focused checks passed after the merge; zero failures/skips. Earlier full-suite measurements remain tied to their named feature commits above. The literal rebase checklist remains unticked because history was merged, not rewritten. Fresh game/Linux checks remain WIP.

TheDevil added 1 commit

2026-11-14 23:45:27 +00:00

fix(vi): pace the 30 FPS cap in the host

PR description / Description follows the template (pull_request_target) Successful in 5s

Details

CI / Conventional Commit title (pull_request) Has been cancelled

Details

CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Has been cancelled

Details

CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Has been cancelled

Details

CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Has been cancelled

Details

CI / Workflow lint (actionlint) (pull_request) Has been cancelled

Details

Third-party parts / As their notes say (pull_request) Has been cancelled

Details

Third-party parts / The same as upstream (pull_request) Has been cancelled

Details

74f1da1924

Assisted-by: OpenAI Codex

TheDevil commented

2026-11-15 00:17:35 +00:00

Author

First-time contributor

Aligned the description with the current template and named fixture results. Pending in-game/Linux validation remains explicit and the PR stays WIP.

TheDevil added 1 commit

2026-11-15 00:41:44 +00:00

Merge remote-tracking branch 'origin/main' into fix/30fps-host-pacing

CI / Conventional Commit title (pull_request) Blocked by required conditions

Details

CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Blocked by required conditions

Details

CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Blocked by required conditions

Details

CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Blocked by required conditions

Details

CI / Workflow lint (actionlint) (pull_request) Blocked by required conditions

Details

Third-party parts / As their notes say (pull_request) Blocked by required conditions

Details

Third-party parts / The same as upstream (pull_request) Blocked by required conditions

Details

PR description / Description follows the template (pull_request_target) Successful in 4s

Details

Website / Website on Cloudflare (pull_request_target) Has been skipped

Details

a9c0551156

TheDevil commented

2026-11-15 00:43:03 +00:00

Author

First-time contributor

Updated the branch to include main @ 6c9edffe by non-rewriting merge. Current head a9c0551156; 5 focused post-merge checks passed, 0 failed/skipped. Description preserves commit-specific full-suite results and remaining live/Linux validation. Still WIP.

Updated the branch to include main @ 6c9edffe by non-rewriting merge. Current head a9c0551156ae2043f441b4cfee05f1784fd58ed5; 5 focused post-merge checks passed, 0 failed/skipped. Description preserves commit-specific full-suite results and remaining live/Linux validation. Still WIP.

TheDevil referenced this pull request

2026-11-15 00:46:56 +00:00

WIP: feat(pc): add KBM actions, item wheels, DLSS 5 HDR and 30 FPS #173

TheDevil referenced this pull request

2026-11-15 00:46:59 +00:00

WIP: feat(pc): add KBM actions, item wheels, DLSS 5 HDR and 30 FPS #173

llun commented

2026-11-15 01:39:03 +00:00

Member

Tested on RTX 5080, Windows 11: pass. 30 holds steadily, game time stays real, and switching between 30 and 60 in play works.

Builds. After: this PR @ a9c05511, built with make_game.py (DLSS, shaders prepared). Before: main @ 25726217 (host/nn_vi.c and gpu/vk/fps_stats.h have not changed between it and 6c9edffe). Same saves and options on both.

Method. Runs were headless with audio off and real pacing (no HOS_NO_PACING, no HOS_REALTIME_STEP), with HOS_FPS_LOG=1 and RX_VK_FRAME_LOG. I stood still in Kokiri Forest after a teleport and changed the Game frame rate cap through the control bridge (option set frame_rate_cap) every 600 presents: saved value → 30 → 60 → 30 → 60, then Game speed Per frame → 30 → 60. Main and the PR were run alternately, 3 runs each. The saved cap was 5 (240) at the start. Frame times are present intervals from the frame log, and the medians are per 600-present segment.

	Before: main @ `25726217`	After: this PR @ `a9c05511`
Lowest choice in the Game frame rate cap row	60 (`option set` -1 is cut to 0)	30
Menu 30, median frame time (runs 1 / 2 / 3)	16.6 ms (it stays at 60)	33.3 / 33.3 / 33.4 ms, 30.00 fps
Menu 30, p1–p99	—	30.3–36.8 ms (runs 1 and 3, both 30 segments)
Back to 60, median	16.6 ms	16.7 / 16.7 / 16.7 ms, 60.00 fps
30 with Game speed Per frame	—	33.3 ms, `vi: game clock` speed 1.000 (real time), and Per frame still saved when back at 60
`HOS_FPS_MAX=30` (2 runs each)	16.7 / 16.7 ms (60)	33.3 / 33.3 ms (30), `vi: frame rate cap 30` from the start
`vi: game clock` speed at 30	—	0.992–1.006, median 1.000 (58 lines over 5 runs)
Saved `frame_rate_cap=5` at start	240	240 (`vi: frame rate cap 240`)
Saved `frame_rate_cap=-1` at start	—	30 from the start, then 30 → 60 → 30 → 60 all correct (1 run)

Each switch logs one vi: frame rate cap 30: pacing, the game's limiter and its time step follow it line, or frame rate cap 60: the game's own limiter and step. The host is the only limiter, and nothing in the renderer sleeps. Saves and options were restored after every run and checked with diff -rq.

One PR run (run 2) had a stretch of about half a minute where the game's CPU side slowed down: the game's own frame rate fell to 21, GPU time was 6.3 ms a frame, and speed was 0.76 for one 5 s window. It started in the middle of a 30 segment and went on through the next 60 one, so it doesn't follow a cap switch. It didn't come back: run 3, with the same steps, was clean, so it was most likely other load on the PC. Run 2's other segments were 33.3 and 16.7 ms.

Our review on #173 (comment 20679), point by point:

Two limiters fighting: fixed. 30 is the host pacer's own period, and present_interval is 1 at 30, so the console's interval 2 doesn't halve it to 15. There is no renderer sleep, and test_frame_rate_cap_30.py checks that none is added.
HOS_FPS_MAX=30 still giving 60: fixed in both fps_of and fps_stats::cap_rate, and seen in game.
HOS_FPS_MAX read every frame: fixed. The host reads it once (the read flag also fixes main's start = -1 sentinel, which clashed with -1), and the renderer reads it once (static).
Game speed forced to real time without saying so: it is now said in both rows' help text, in the readme and in development.md. The saved Game speed is kept and used again at 60, which I saw in game.

Nit (not blocking): the options file loads a value under -1 as -1, since it is clamped to the row's lo, so a hand-edited frame_rate_cap=-2 gives 30 once the renderer is up. The fixture's (-2, unset) → 60 case doesn't include that clamp. No build writes such a value.

Game-free suite (python -I -B tests/run_all.py, llvm-mingw): this PR 523 pass / 16 fail / 42 skip; main 6c9edffe 522 / 16 / 42 with the same 16 failures (appimage_player_build, bake_portability, cpm_discovery, crash_dialog, env_on, finish_output, lift_cache, lift_cleanup_error, linux_preflight, make_game_cache, oot_state_tap, readback_late, tested_configurations, vk_pgo, vram_budget_row, walk_key). The extra pass is the new test_frame_rate_cap_30.py.

Not tested here: a cutscene, audio (it was off), a real display with V-Sync (the runs were headless), frame generation, HOS_GAME_INTERVAL=1, and Linux.

Tested on RTX 5080, Windows 11: **pass**. 30 holds steadily, game time stays real, and switching between 30 and 60 in play works. **Builds.** After: this PR @ a9c05511, built with `make_game.py` (DLSS, shaders prepared). Before: main @ 25726217 (`host/nn_vi.c` and `gpu/vk/fps_stats.h` have not changed between it and 6c9edffe). Same saves and options on both. **Method.** Runs were headless with audio off and real pacing (no `HOS_NO_PACING`, no `HOS_REALTIME_STEP`), with `HOS_FPS_LOG=1` and `RX_VK_FRAME_LOG`. I stood still in Kokiri Forest after a teleport and changed the Game frame rate cap through the control bridge (`option set frame_rate_cap`) every 600 presents: saved value → 30 → 60 → 30 → 60, then Game speed Per frame → 30 → 60. Main and the PR were run alternately, 3 runs each. The saved cap was 5 (240) at the start. Frame times are present intervals from the frame log, and the medians are per 600-present segment. | | Before: main @ 25726217 | After: this PR @ a9c05511 | |---|---|---| | Lowest choice in the Game frame rate cap row | 60 (`option set` -1 is cut to 0) | 30 | | Menu 30, median frame time (runs 1 / 2 / 3) | 16.6 ms (it stays at 60) | 33.3 / 33.3 / 33.4 ms, 30.00 fps | | Menu 30, p1–p99 | — | 30.3–36.8 ms (runs 1 and 3, both 30 segments) | | Back to 60, median | 16.6 ms | 16.7 / 16.7 / 16.7 ms, 60.00 fps | | 30 with Game speed Per frame | — | 33.3 ms, `vi: game clock` speed 1.000 (real time), and Per frame still saved when back at 60 | | `HOS_FPS_MAX=30` (2 runs each) | 16.7 / 16.7 ms (60) | 33.3 / 33.3 ms (30), `vi: frame rate cap 30` from the start | | `vi: game clock` speed at 30 | — | 0.992–1.006, median 1.000 (58 lines over 5 runs) | | Saved `frame_rate_cap=5` at start | 240 | 240 (`vi: frame rate cap 240`) | | Saved `frame_rate_cap=-1` at start | — | 30 from the start, then 30 → 60 → 30 → 60 all correct (1 run) | Each switch logs one `vi: frame rate cap 30: pacing, the game's limiter and its time step follow it` line, or `frame rate cap 60: the game's own limiter and step`. The host is the only limiter, and nothing in the renderer sleeps. Saves and options were restored after every run and checked with `diff -rq`. One PR run (run 2) had a stretch of about half a minute where the game's CPU side slowed down: the game's own frame rate fell to 21, GPU time was 6.3 ms a frame, and speed was 0.76 for one 5 s window. It started in the middle of a 30 segment and went on through the next 60 one, so it doesn't follow a cap switch. It didn't come back: run 3, with the same steps, was clean, so it was most likely other load on the PC. Run 2's other segments were 33.3 and 16.7 ms. **Our review on #173 (comment 20679), point by point:** - Two limiters fighting: fixed. 30 is the host pacer's own period, and `present_interval` is 1 at 30, so the console's interval 2 doesn't halve it to 15. There is no renderer sleep, and `test_frame_rate_cap_30.py` checks that none is added. - `HOS_FPS_MAX=30` still giving 60: fixed in both `fps_of` and `fps_stats::cap_rate`, and seen in game. - `HOS_FPS_MAX` read every frame: fixed. The host reads it once (the `read` flag also fixes main's `start = -1` sentinel, which clashed with -1), and the renderer reads it once (`static`). - Game speed forced to real time without saying so: it is now said in both rows' help text, in the readme and in development.md. The saved Game speed is kept and used again at 60, which I saw in game. Nit (not blocking): the options file loads a value under -1 as -1, since it is clamped to the row's `lo`, so a hand-edited `frame_rate_cap=-2` gives 30 once the renderer is up. The fixture's `(-2, unset) → 60` case doesn't include that clamp. No build writes such a value. **Game-free suite** (`python -I -B tests/run_all.py`, llvm-mingw): this PR 523 pass / 16 fail / 42 skip; main 6c9edffe 522 / 16 / 42 with the same 16 failures (appimage_player_build, bake_portability, cpm_discovery, crash_dialog, env_on, finish_output, lift_cache, lift_cleanup_error, linux_preflight, make_game_cache, oot_state_tap, readback_late, tested_configurations, vk_pgo, vram_budget_row, walk_key). The extra pass is the new `test_frame_rate_cap_30.py`. **Not tested here:** a cutscene, audio (it was off), a real display with V-Sync (the runs were headless), frame generation, `HOS_GAME_INTERVAL=1`, and Linux.

mooki added the

labels

2026-11-15 13:56:16 +00:00

HumanBean commented

2026-11-15 18:47:08 +00:00

Owner

This pull request now conflicts with main: host/nn_vi.c, which #488 changed (the movie pacing). Could you rebase it onto the current main? Thank you.

This pull request now conflicts with `main`: host/nn_vi.c, which https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/488 changed (the movie pacing). Could you rebase it onto the current `main`? Thank you.

CI / Conventional Commit title (pull_request) Blocked by required conditions

Details

CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Blocked by required conditions

Details

CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Blocked by required conditions

Details

CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Blocked by required conditions

Details

CI / Workflow lint (actionlint) (pull_request) Blocked by required conditions

Details

Third-party parts / As their notes say (pull_request) Blocked by required conditions

Details

Third-party parts / The same as upstream (pull_request) Blocked by required conditions

Details

PR description / Description follows the template (pull_request_target) Successful in 4s

Details

Website / Website on Cloudflare (pull_request_target) Has been skipped

Details

This pull request has changes conflicting with the target branch.

docs/development.md
host/nn_vi.c

Some workflows are waiting to be reviewed.

View command line instructions

Manual merge helper

Use this merge commit message when completing the merge manually.

Merge commit title

WIP: fix(vi): pace the 30 FPS cap in the host (#609)

Merge commit body

Reviewed-on: https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/609

WIP: fix(vi): pace the 30 FPS cap in the host (#609)

Reviewed-on: https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/609

Checkout

From your project repository, check out a new branch and test the changes.

git fetch -u 'fix/30fps-host-pacing:TheDevil-fix/30fps-host-pacing'

git switch 'TheDevil-fix/30fps-host-pacing'

Rows
Columns

WIP: fix(vi): pace the 30 FPS cap in the host #609

Why

How to reproduce

What changed

Measurements

Flags and settings

Testing

Risks

Checklist

Everyone

When it applies

Upstream refresh

Manual merge helper

Checkout